Skip to content

Latest commit

 

History

History
86 lines (54 loc) · 3.37 KB

dev_tools.md

File metadata and controls

86 lines (54 loc) · 3.37 KB
Raw

Local development tools

Running tests

With Studio's services running, you may run tests with the following commands:

# backend
make test
# frontend
yarn run test

View more testing tips

Linting

Front-end linting is run using:

yarn run lint-frontend

Some linting errors can be fixed automatically by running:

yarn run lint-frontend:format

Make sure you've set up pre-commit hooks as described above. This will ensure that linting is automatically run on staged changes before every commit.

Profiling and local production testing

If you want to test the performance of your changes, you can start up a local server with settings closer to a production environment like so:

# build frontend dependencies
yarn run build
# run the server (no webpack)
yarn run runserver
# or for profiling production more closely
yarn run runserver:prod-profiling

Once the local production server is running, you can also use Locust to test your changes under scenarios of high demand like so:

cd deploy/chaos/loadtest
make timed_run
make stop_slaves  # mac: killall python

Profiling

In case you need to profile the application to know which part of the code are more time consuming, there are two different profilers available to work in two different modes. Both will store the profiling output in a directory that's determined by the PROFILE_DIR env variable. If this variable is not set, the output files will be store in a folder called profiler inside the OS temp folder (/tmp/profile usually) Note that both profiling modes are incompatible: you can either use one or the other, but not both at the same time. In case the env variables are set for both modes, All request profiling mode will be used.

All requests profiling mode

This mode will create interactive html files with all the profiling information for every request the Studio server receives. The name of the files will contain the total execution time, the endpoint name and a timestamp.

To activate it an env variable called PROFILE_STUDIO_FULL must be set.

Example of use:

PROFILE_STUDIO_FULL=y yarn runserver

Afterwards no further treatment of the generated files is needed. You can open directly the html files in your browser.

Endpoint profiling mode

When using the all requests mode it's usual that the profile folder is soon full of information for requests that are not interesting for the developer, obscuring the files for specific endpoints.

If an env variable called PROFILE_STUDIO_FILTER is used, the profiler will be executed only on the http requests containing the text stated by the variable.

Example of use:

PROFILE_STUDIO_FILTER=edit yarn localprodserver

For this case, only html requests having the text edit in their request path will be profiled. The profile folder will not have html files, but binary dump files (with the timestamp as filename) of the profiler information that can be later seen by different profiling tools (snakeviz that can be installed using pip is recommended). Also while the server is running, the ten most time consuming lines of code of the filtered request will be shown in the console where Studio has been launched.

Example of snakeviz use:

snakeviz /tmp/profile/studio\:20200909161405011678.prof

will open the browser with an interactive diagram with all the profiling information